SDN-Based Application-Aware Segment Routing for Large-Scale Network

With the rapid growth of fifth-generation (5G), network operators have met with difficulties in ensuring user satisfaction due to diverse service-level agreement (SLA) requirements. Application-aware routing is able to potentially address this issue in implementing differentiated routing policies corresponding to various applications. To facilitate the application-aware routing, software-defined networking (SDN) and segment routing (SR) are potential solutions with programmability property to steer network traffic into appropriate routing paths. However, application-aware SR encounters two main problems, including application identification for encrypted traffic and path identification with human intervention. Therefore, this article proposes a new SDN-based SR mechanism that could help network operators overcome these problems. This approach implements reinforcement learning to adapt to dynamic networks in order to optimize the quality of experience (QoE) that network operators must guarantee. The proposal also considers the application class to implement corresponding routing policies for various applications to meet the stringent SLA requirements. Identifying the applications is sometimes complicated due to encrypted network traffic. Hence, our proposal implements a traffic classification approach to classify encrypted traffic into different kinds of applications. Obtained results under considered conditions illustrate that our proposal outperforms the standard SR mechanism related to QoE and decreases up to 64.39% of overhead compared to several benchmarks.